Pollster.com: Gore in NH: 12% or 32%?

June 28, 2007

The curious poll result du jour: As we just posted, a new 7 News/ Suffolk University poll of New Hampshire voters shows Hillary Clinton leading the Democratic primary contest (with 37%) vote with Barack Obama (19%) and the rest of the pack trailing far behind. Yet according to the release:

Twenty-nine percent of Clinton voters would switch to Gore if he announced for president, and when all of the switches from other Democratic candidates were recalculated, Gore would defeat Clinton. In total, 32 percent of Democratic voters would support Gore over the candidate they are currently leaning toward.

That result conflicts with other recent New Hampshire surveys that include Gore in their trial heat vote question and show him with nowhere near that level of support. The recent CNN/WMUR/University of New Hampshire survey had Gore at 12%. Other New Hampshire polls conducted since March showed Gores support varying between 8% and 15%. So what's up with this new result?

The answer is almost certainly in the very different ways these pollsters measure support for Gore. Let's start with the two-part question asked by the CNN/WMUR/UNH poll, which is also the approach taken by most other pollsters:

Q7. I'm going to read you the names of the candidates who are either running or considering running for the Democratic nomination. If the Democratic primary for president were held today, which of the following would you support for the Democratic nomination -- Joe Biden, Hillary Clinton, Chris Dodd, John Edwards, Al Gore, Mike Gravel, Dennis Kucinich, Barack Obama, Bill Richardson or someone else?

Q8. Who is your second choice?

Again, on this survey conducted June 6-10, only 12% support Gore, while Clinton leads with 36%, to 22% for Obama, 12% for Edwards, 10% Richardson and all other candidates in the low single digits. They recalculated using the second choices of Gore voters, and showed Clinton receiving 39%, Obama 24%, Edwards 11% and Richardson 10% without Gore in the race.

Now consider the very different two-part question taken by the 7 News/ Suffolk University poll (PDF):

Q14. The following eight Democratic candidates may seek the Democratic nomination. Listed alphabetically on your ballot, they are: Joe Biden, Hillary Clinton, Christopher Dodd, John Edwards, Mike Gravel, Dennis Kucinich, Barack Obama, Bill Richardson... For whom would you vote or toward whom would you lean at this time?

Q15. If Al Gore were to enter the presidential race, would you support him over the candidate you are currently leaning towards?

Clinton's lead over (37% to 19% for Obama) on the first 7 News/Suffolk University question is very similar to her lead (39% to 24%) on the recalculated vote without Gore on the CNN/WMUR/UNH poll. Yet despite the relatively small sample size of Democrats in each poll (n=232 for the former, n=309 for the latter), the difference in Gore's support is highly statistically significant.

So while some other minor variation in methodology may explain some of the disparity, the most likely culprit is the very different way the pollsters measure Al Gore's potential support:

CNN/WMUR/UNH and most other pollsters include Gore on the laundry list of "potential" Democrats. Given the results, it seems that some voters who might support Gore are discounting the possibility that he will run and tentatively supporting for other candidates.

7 News/Suffolk University is taking a deeper step into what CBS News pollster Kathy Frankovic calls "Pollster Land," by asking respondents to imagine that Al Gore has entered the presidential race. Perhaps more important, rather than repeating the list of choices with Gore included, they simply pose a "yes/no" question, "would you support him over the candidate you are currently leaning towards?" The result is probably a bit of what pollsters call "acquiescence bias," a tendency of some respondents to agree with all questions.

So which is the better approach? Neither, or perhaps both. The reality for New Hampshire voters is probably somewhere in between these two results. Since many well informed voters have likely concluded Gore will not run, the trial heat questions asked by most pollsters understate his potential support. On the other hand, the 7 News/Suffolk University poll probably overstates it. In either event, the way an actual Gore candidacy might alter current preferences is hypothetical and difficult to predict.

Keep in mind what is probably the most important result from the CNN/WMUR poll: Nearly half of New Hampshire Democrats (49%) say they have "no idea" who they will vote for in the presidential primary, while only 8% have "definitely decided" on a candidate. So the answers many respondents give to these early trial heat questions are tentative, at best.

Finally, the fact that some pollsters include Gore in their trial heat questions while others do not raises an important issue: Which should we include on our charts here are Pollster.com? My colleague Charles Franklin addresses that issue today in a separate post.

typo corrected

-- Mark Blumenthal

June 28, 2007 in The 2008 Race